Process Simulation of Isoprene Recovery from Dilute Fermentation Off-Gas Using Absorption–Desorption

Abstract

The recovery of biobased isoprene from dilute fermentation off-gas remains a major challenge due to low product concentrations and energy-intensive downstream separations. In this study, a process intensification strategy based on absorption–desorption was developed and evaluated to efficiently recover isoprene from air mixtures. The proposed system integrates absorption using 3-methyltridecane, vacuum degassing, steam stripping, and condensation with multiple recycle loops for solvent, water, and residual isoprene. Of the thermodynamic models examined, the Peng–Robinson equation of state was used for the absorption section while UNIFAC was used for steam stripping. Process optimization demonstrated up to 95% isoprene recovery and 98% product purity while significantly reducing energy consumption. Sensitivity analyses revealed that solvent temperature, solvent-to-feed ratio, and feed isoprene concentration strongly influence process efficiency, with energy demand decreasing from 26 to 10 MJ kg^–1 as feed concentration increased. Compared to literature benchmarks, the proposed design achieved up to 40% reduction in steam usage. These results highlight absorption-based separation as a viable and energy-efficient pathway for recovering volatile hydrocarbons from dilute bioprocess streams and provide design insights for future scale-up and experimental validation.

1. Introduction

The growing global demand for renewable alternatives to fossil-derived products has driven the development of biorefineries. These are advanced facilities that convert biomass into an array of biofuels, biochemicals, and biomaterials. Unlike conventional petroleum refineries, biorefineries use renewable feedstocks such as lignocellulosic biomass, agricultural residues, and microalgae to produce high-value compounds, reducing greenhouse gas emissions. This shift from fossil-based petroleum refineries to renewable biorefineries has positioned biorefineries as cornerstones of the green industrial revolution, with projections suggesting that biobased chemicals could account for over 20% of the global chemical market by 2030. Among these chemicals, isoprene is a volatile C5 hydrocarbon, notable for its widespread use in manufacturing synthetic rubber, adhesives, and specialty polymers. Its importance lies in its role as a precursor to polyisoprene, which provides the elasticity and durability required for high-performance applications. Traditionally, isoprene has been produced from fossil-based processes such as naphtha cracking or as a byproduct of ethylene production, both of which are energy-intensive and dependent on finite resources. The emerging transition to biobased isoprene production within biorefineries not only aligns with sustainability goals but also mitigates supply chain vulnerabilities tied to fossil feedstocks.

Bioprocess routes employ microbial or enzymatic synthesis pathways using renewable feedstocks like glucose or biomass. While these methods reduce dependence on fossil carbon, their overall energy demand can vary. For example, in microbial synthesis, fermentation itself is generally less energy-intensive than petrochemical cracking, but downstream separations often remain costly. These bioprocess routes typically leverage engineered strains of microorganisms to produce isoprene through the mevalonate (MVA) or methylerythritol phosphate (MEP) pathways.

The shift toward biobased isoprene is particularly significant in mitigating environmental impacts associated with traditional petrochemical methods. Research demonstrates that biobased polyisoprene, derived from renewable resources like agricultural residues, offers a lower carbon footprint and reduced land use compared to natural and synthetic rubber, aligning with sustainability goals. Additionally, advancements in catalysts and biotechnological methods have enhanced the efficiency of isoprene production from biobased feedstocks, making it a viable alternative to petroleum-derived processes. Companies like Goodyear are also investing in biobased isoprene for tire production, highlighting its industrial and commercial potential.

Following production, recovering isoprene from biobased fermentation presents significant downstream separation challenges because the crude product stream contains water, residual sugars, alcohols, and other volatile impurities, all of which necessitate robust purification steps to achieve market-ready quality. Another barrier in biobased isoprene production is the extremely low concentration of isoprene in feed gas. , Reported titers can be as low as 0.0009 g L^–1 in aerobic fermentation and 0.0021 g L^–1 in fed-batch fermentation. These low yields arise from inherent metabolic and pathway limitations, including restricted flux through biosynthetic routes and inefficiencies in enzyme activity or cofactor availability. Isoprene concentration in the off-gas depends primarily on microbial productivity and oxygen flow rate. Attempting to increase titers is further constrained by oxygen limitation in air. Increasing the oxygen flow rate would require excess oxygen and substantially raise fermentation costs.

Several downstream separation strategies have been explored for recovering volatile hydrocarbons such as isoprene from biobased systems, each involving trade-offs in energy demand, selectivity, scalability, and cost. Conventional and emerging methods include liquid–liquid extraction, gas stripping, membrane separation, and integrated in situ approaches such as pervaporation or gas stripping during fermentation. Those have been investigated to enhance hydrocarbon recovery and productivity from dilute biostreams. , Distillation, including steam-distillation, exploits boiling point differences and has been applied to aqueous biomixtures containing volatile hydrocarbons; however, its high energy intensity limits sustainability for large-scale deployment. , Lower-energy alternatives include adsorption-based separations using porous materials such as zeolites, metal–organic frameworks (MOFs), and covalent organic frameworks (COFs), which enable selective hydrocarbon capture through size exclusion or specific interactions, exemplified by zeolite 5A for shape-selective separations, Fe-MOF-74 for unsaturated hydrocarbons, and COFs exploiting π–π interactions in complex gas mixtures. Membrane-based processes offer further energy savings, particularly for dilute fermentation streams, with polymeric membranes such as PDMS enabling selective pervaporation of hydrophobic organics and hybrid membrane–distillation systems achieving substantial reductions in energy consumption. , Solvent-based absorption and extraction methods remain effective for low-volatility or trace hydrocarbons, using nonpolar solvents to achieve high recovery with limited coextraction, as demonstrated in biofuel processing and in patented systems combining solvent extraction with steam stripping for isoprene recovery from fermenter off-gas. Emerging biosorption approaches using microbial biomass have also shown potential for hydrocarbon removal from aqueous systems, including fungal consortia capable of degrading polycyclic aromatic hydrocarbons.

While various separation technologies have been proposed, each presents trade-offs in terms of energy use, scalability, and cost. Among these, absorption offers a promising pathway for recovering volatile hydrocarbons like isoprene, particularly when integrated with solvent regeneration and recycling. Its relatively low energy intensity and operational simplicity make it an attractive option for large-scale processing of dilute hydrocarbon streams. However, no work has focused on systematically studying absorption parameters such as solvent-to-feed ratio and solvent temperature, which can reduce energy intensity while maintaining high recovery. This leaves a critical gap for process optimization, motivating the present study to determine the viability of the absorption system for efficient isoprene extraction. In this study, we present a comprehensive absorption process simulation using Aspen Plus V14 for the purification of isoprene from a dilute air mixture, focusing on achieving a liquid-phase product. Our approach involved designing and optimizing a separation process that effectively isolates isoprene while minimizing energy consumption. Key analyses were conducted to evaluate process efficiency, including sensitivity studies on critical parameters. The findings from this work provide valuable insights into the viability of isoprene purification through simulation-based process design and lay the groundwork for future experimental validation and industrial implementation.

2. Methods

2.1. Process Design and Simulation

In our paper, we first designed a process that was able to extract isoprene from the fermenter off-gas into a liquid form. Additional components were added to the process to make the purity of isoprene higher in the final product. Different parameters in the process were analyzed using Aspen Plus sensitivity tools to keep the energy requirements as low as possible while maintaining high recovery and purity of isoprene. Figure illustrates the process diagram of the integrated absorption and desorption system. The process consists of a 32-stage absorber and a 35-stage stripper column, both operating with 3-methyltridecane as the absorption solvent. 3-Methyltridecane was selected as a representative heavy, branched C₁₄ hydrocarbon whose properties suit this separation: its very low volatility minimizes solvent carryover into the treated gas and product and allows easy regeneration by steam stripping, while its nonpolar character gives good affinity for isoprene, keeps the solubility of the permanent gases (air and CO₂) low, and provides near-complete immiscibility with water for clean water–solvent separation downstream. A pressure drop of 0.5 bar is included in the absorber, which processes a large gas throughput (∼1500 kg h^–1). The stripper carries a much smaller vapor (steam) flow and therefore has a comparatively negligible pressure drop, which was neglected; it was modeled at constant 1 bar. For the base case, the absorber operates with cold lean solvent fed at 2 °C and feed gas entering at 40 °C while the stripper operates near the steam temperature of ≈104 °C. The absorber top stage is at 1.5 bar; with the 0.5 bar column pressure drop the bottom is at 2 bar, so the treated gas leaves the top at 1.5 bar and the rich solvent leaves the bottom at 2 bar.

1.

Process flow diagram of the absorption process in Aspen Plus. The boxed numbers refer to the stream numbers.

The streams in Table were simulated for 1500 kg h^–1 feed gas with 1 wt % isoprene concentration. This flow rate and composition define the base case. The input conditions for each simulation case in this study are summarized in Table .

1. Condition and Composition of Main Streams of the Process.

				mass fraction
stream	mass flow rate (kg h–1)	temperature (°C)	pressure (bar)	air	isoprene	3-methyltridecane	water	CO2
1	1500	40.0	2	9.757 × 10–1	1.011 × 10–2	4.160 × 10–11	4.251 × 10–7	1.422 × 10–2
2	1801	1.0	1.5	4.518 × 10–9	3.094 × 10–4	9.997 × 10–1	0	2.919 × 10–8
3	1483	3	1.5	9.855 × 10–1	2.549 × 10–4	4.909 × 10–6	2.415 × 10–7	1.419 × 10–2
3A	1483	13.0	1.5	9.855 × 10–1	2.549 × 10–4	4.909 × 10–6	2.415 × 10–7	1.419 × 10–2
4	1818	14.6	2.5	1.028 × 10–3	8.443 × 10–3	9.904 × 10–1	1.538 × 10–7	1.526 × 10–4
5	2.337	1.0	0.1	7.715 × 10–1	1.497 × 10–1	1.161 × 10–4	5.633 × 10–6	7.868 × 10–2
6	1816	104.0	1	3.628 × 10–5	8.261 × 10–3	9.917 × 10–1	1.467 × 10–7	5.149 × 10–5
7	24.30	104.0	1	2.701 × 10–9	2.056 × 10–8	2.084 × 10–14	1.000 × 100	1.315 × 10–6
8	40.02	104.2	1	1.646 × 10–3	3.643 × 10–1	5.160 × 10–2	5.801 × 10–1	2.336 × 10–3
9	1800	103.0	1	4.427 × 10–13	2.327 × 10–4	9.992 × 10–1	6.018 × 10–4	8.404 × 10–10
12	16.88	90.0	1	3.902 × 10–3	8.637 × 10–1	1.223 × 10–1	4.552 × 10–3	5.537 × 10–3
13	23.14	21.0	1	2.836 × 10–9	2.159 × 10–8	2.189 × 10–14	1.000 × 100	1.381 × 10–6
14	14.88	100.0	1	4.426 × 10–3	9.704 × 10–1	1.379 × 10–2	5.136 × 10–3	6.278 × 10–3
15	1.999	100.0	1	4.070 × 10–6	6.926 × 10–2	9.305 × 10–1	2.019 × 10–4	2.555 × 10–5
16	0.2552	15.0	1	2.008 × 10–1	6.638 × 10–1	2.446 × 10–7	2.499 × 10–3	1.329 × 10–1
17	14.63	15.0	1	9.994 × 10–4	9.757 × 10–1	1.404 × 10–2	5.182 × 10–3	4.069 × 10–3
P-1	1.084	23.9	1	0	0	0	1	0
P-2	0.006120	50.9	2	0	0	0	0	1

2. Summary of Simulation Cases and Input Conditions for Each figure.

figure	parameter varied	feed gas mass flow rate, kg h–1	isoprene feed concentration, wt %	solvent to feed gas ratio	Solvent temperature, °C	vacuum degasser pressure, atm	steam mass flow rate, kg h–1
2	thermodynamic model	1500	1	1.2	2	0.1	24.3
3	thermodynamic model	1500	1	1.2	2	0.1	24.3
4	solvent temperature	1500	1	1.2	–10 to 35	0.1	24.3
5	feed gas isoprene concentration	1500	1–4.3	1.2	2	0.1	24.3
6	feed gas isoprene concentration	1500	1–4.3	1.2	2	0.1	24.3
7	solvent to feed gas ratio	800/1500/3000	1	0.8–3	2	0.1	8.2–124
9	vacuum degasser pressure	1500	1	1.2	2	0.1–1.5	24.3
10	vacuum degasser pressure	1500	1	1.2	2	0.1–1.5	24.3
11	(base case)	1500	1	1.2	2	0.1	24.3

The feed gas (stream-0), comprising air, isoprene, and CO₂, enters the system at 40 °C and 2 bar. Residual isoprene from the condenser (S-4) is combined with fresh feed gas and introduced into the absorber (T-1), while the solvent (stream-2) is fed from the top of the absorber. We used the RadFrac column in Aspen Plus for modeling the stripper column and absorber column. Within the absorber, 99.86% of air and 98% of CO₂ are effectively removed from the gas phase (stream-3), significantly reducing noncondensable components. Stream-3 is used to cool the solvent going to the absorber using a heat exchanger (HE-3). The isoprene-rich solvent (stream-4) is then directed to a vacuum degasser (S-1), where, under 0.1 atm pressure, air and CO₂ are further removed. The degassed stream (stream-6) is subsequently passed through a pump, heat exchanger, and heater before being introduced at the top of the steam stripper column (T-2). Simultaneously, steam (stream-7) is introduced at the bottom in a counter-current configuration. Within the stripper, 97% of the isoprene is desorbed into the steam phase, while over 99% of the solvent (stream-9) is regenerated and returned to the absorber solvent tank (R-1). It is critical that the isoprene-rich solvent is preheated to match the steam temperature, as cooler solvent temperatures may condense the steam and reduce separation efficiency. From the top of the stripper, the isoprene-water mixture (stream-8) is routed to a flash separator (S-2), where the stream is cooled to 21 °C to facilitate water-isoprene separation. Approximately 95% of the water is recycled back (stream-13) to the stripper column. The isoprene stream (stream-12) is then sent to a secondary flash separator (S-3) to remove residual solvent, which is subsequently recycled for reuse. The purified isoprene (stream-14) is finally condensed in a condenser (S-4), where 98.8% is recovered as liquid at 15 °C, with the remaining vapor-phase isoprene (stream-16) recycled to the feed gas.

Three distinct recycle loops are incorporated: the solvent recycle loop (stream-10), the water recycle loop (stream-13), and the residual isoprene recycle loop (stream-16). For the solvent and water recycling loops, a fixed flow rate has been established, and a makeup stream is employed to make up for losses of solvent and water that occur during the process. Additionally, the residual isoprene recycle loop (stream-16) is implemented to loop back the isoprene that escaped into the gas phase to the feed, thereby enhancing the overall efficiency of the process and reducing material losses. Two small purge streams prevent water and CO₂ from accumulating in the recycle loops. The P-1 stream (∼1.08 kg h^–1) corresponds to decantation of the aqueous phase, which separates from the hydrocarbon solvent owing to the very low mutual solubility of water and 3-methyltridecane. P-2 stream (∼0.006 kg h^–1) removes CO₂ that accumulates in the gas recycle loop. Both streams contain negligible isoprene and solvent (<0.01% of throughput), so the recovery, purity, and energy results are unaffected.

2.2. Process Performance Metrics

Two metrics are used to evaluate the process: the absorber efficiency and the energy requirement per unit of isoprene produced.

The absorber efficiency, η_abs, quantifies the fraction of isoprene captured from the gas phase. It is defined as the isoprene in the bottom liquid product ( ${ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{b}\mathrm{p}})$ of the absorber relative to the isoprene going into the absorber ( ${ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{i}\mathrm{n}})$ .

$${\eta }_{\mathrm{a}\mathrm{b}\mathrm{s}}=\frac{{ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{b}\mathrm{p}}}{{ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{i}\mathrm{n}}}$$ 1

where ${ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{i}\mathrm{n}}$ & ${ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o},\mathrm{b}\mathrm{p}}$ are the isoprene mass flow rates in the gas going into the absorber and the liquid from the bottom of the absorber.

The energy requirement, E, per kg of isoprene is the electrical energy consumed by the process per kilogram of isoprene produced.

$$E=\frac{\sum _i{E}_i}{{ṁ}_{\mathrm{i}\mathrm{s}\mathrm{o}}}$$ 2

where E _i is the electrical energy demand of unit i and ṁ_iso is the liquid-isoprene product rate. The pumps, compressor, and vacuum system are counted as their direct electrical (shaft) work, and the electric heaters as their electrical duty. The cooling and refrigeration loads are not counted as raw thermal duties; instead, each cooling duty is converted to the electrical work. Because the cooling duties are to subambient temperatures (the solvent chiller at 1–2 °C and the condenser at 15 °C), they are provided by mechanical refrigeration rather than cooling water. Expressing every term as electrical energy places heating and cooling on a common, additive basis, rather than summing thermal duties of differing grade. The stripping steam is slightly superheated at 104 °C and 1 bar. Even though it is reported separately on a mass basis (kg steam per kg isoprene; Figures and ), the electrical energy to generate it is accounted for in the E term.

6.

Effect of isoprene concentration in the feed gas on solvent and steam requirements per kilogram of isoprene produced (simulation conditions in Table ).

7.

Effect of solvent-to-feed gas ratio on steam usage at three different feed gas flow rates (simulation conditions in Table ).

2.3. Thermodynamic Model

2.3.1. Absorption

Modeling the absorption of isoprene from a gas mixture into 3-methyltridecane solvent in Aspen Plus requires an appropriate thermodynamic model. Due to limited availability of experimental vapor–liquid equilibrium data for this multicomponent system, model selection was based on expected phase behavior and standard practice in hydrocarbon absorption simulations. Model selection followed a phase-behavior rationale rather than a single global default. , Because the absorber operates at low pressure (≈1.5–2 bar), the vapor phase is essentially ideal, so the controlling nonidealities reside in the liquid phasespecifically the solubilities of the gases and of dilute isoprene in the solvent. Two formulations are possible: a symmetric (φ–φ) equation-of-state approach, in which a single equation of state describes both phases, and an asymmetric (γ–φ) approach, in which a liquid activity-coefficient model is paired with Henry’s law for the noncondensable gases. , Because the absorber fluid is predominantly nonpolar, the symmetric equation-of-state formulation was adopted: a cubic equation of state represents air, CO₂, isoprene, and the solvent within one consistent framework and treats the supercritical permanent gases directly, so an explicit Henry’s-law description is not required. Henry’s law would be necessary only if the dilute gases were significant and the liquid were modeled with activity coefficients, a situation that does not arise in this work. Our absorption system consists of a primarily nonpolar gas mixture (air, isoprene, and CO₂) contacted with a heavy hydrocarbon solvent, 3-methyltridecane. Because the dominant interactions are weak van der Waals forces rather than strong polar or hydrogen-bonding effects, the choice of thermodynamic model is critical for accurately predicting the absorption rate and phase behavior. Various equations of state (EoS) were evaluated, including Soave–Redlich–Kwong (SRK), Peng–Robinson (PENG-ROB), Peng–Robinson with Wong-Sandler mixing rules (PRWS), Universal Functional Activity Coefficient (UNIFAC), Predictive Soave–Redlich–Kwong (PSRK) and non-random two-liquid (NRTL) (Figure ). Each EoS predicted different solubilities for air, CO₂, and isoprene in the solvent, with some variations primarily in air and CO₂ absorption. While all EoS predicted nearly identical isoprene solubility, the main differences arose in the solubility predictions for air and CO₂. Air solubility into the solvent varied significantly, with some models predicting higher absorption than expected, while others predicted low solubility consistent with expected behavior in the solvent. Similarly, CO₂ solubility predictions differed. Differences between EoS may influence predicted recovery and energy demand of isoprene.

2.

Absorber performance to separate isoprene, CO₂ and air from feed gas under different EoS. All the results were obtained under the same feed gas flow rate and the same component composition (simulation conditions in Table ).

PRWS is designed to handle asymmetric mixtures, combining Peng–Robinson EoS for vapor-phase fugacity with an activity coefficient model for liquid-phase nonidealities. It is effective for polar and nonpolar mixtures and provides accurate CO₂ solubility predictions in many cases. However, PRWS overestimated air solubility in the solvent. PRWS is best suited for polar–nonpolar interactions, but our process primarily involves nonpolar species. The complexity in parameter tuning can introduce additional uncertainties in each EoS. While PRWS performed well for CO₂ solubility, its excessive prediction of air solubility makes it unsuitable.

UNIFAC and NRTL work well for highly nonideal liquid mixtures but struggle with permanent gases like air. UNIFAC overpredicted air absorption, indicating misrepresentation of noninteracting gases in solvent, a weakness consistent with its focus on liquid-phase nonidealities rather than gas–liquid equilibria, as outlined by Fredenslund et al. The NRTL model is specifically designed for strong liquid-phase interactions, not weak van der Waals interactions like air and hydrocarbons. Gas-phase behavior is not well modeled, leading to inconsistencies in predicting phase equilibria. Both methods had weaknesses in handling nonpolar gas–liquid systems, making them unsuitable.

Among the evaluated models, the Peng–Robinson (PR) equation of state − was selected for the absorption section. PR is a cubic equation of state extensively validated for nonpolar hydrocarbon and gas-processing systems; at the low absorber pressure it reduces to near-ideal vapor behavior while representing the weakly nonideal hydrocarbon liquid through its standard mixing rules. A practical advantage of the equation-of-state route here is that the supercritical permanent gases (air and CO₂) are represented directly within the same model, without the separate Henry’s-law constants that an activity-coefficient formulation would require. PR produced physically reasonable, low air and CO₂ solubilitiesconsistent with the expected behavior of permanent gases in a heavy hydrocarbon, and in contrast to the overprediction given by the activity-coefficient methods (Figure ). A fully predictive equation of state such as PSRK , was judged unnecessary, since it targets high-pressure systems and vapor-phase nonidealities that are negligible here. Because system-specific binary parameters were not regressed, the absolute gas solubilities carry some uncertainty; however, the screening in Figure confirms that PR reproduces the physically expected trends, supporting its use for this nonpolar absorber.

2.3.2. Desorption

After the isoprene is absorbed into the solvent, it needs to desorb in order to be extracted. Steam stripping is a viable method to do so. In a steam stripping column designed to remove isoprene from a heavy 3-methyltridecane solvent using superheated steam, the choice of the thermodynamic model influences the predicted phase behavior, particularly in systems involving water and hydrocarbons (Figure ). As in the absorber, the stripper operates near atmospheric pressure, so the vapor phase is again close to ideal. By this stage the permanent gases (air and CO₂) have largely been removed in the absorber overhead and the vacuum degasser, so the stripper processes an essentially condensable mixture of water, isoprene, and solvent; the modeling emphasis therefore falls entirely on the strongly nonideal, partially miscible water–hydrocarbon liquid, and no Henry’s-law treatment of dissolved gases is required here either.

3.

Steam stripper column’s performance to separate different components from isoprene-rich solvent into steam under different EoS (simulation conditions in Table ).

Cubic equations of state (EoS), such as Peng–Robinson (PR) and Soave–Redlich–Kwong (SRK), provide a consistent framework for vapor–liquid equilibrium calculations and are widely used for hydrocarbon systems. However, these models have known limitations in representing highly nonideal interactions involving polar components such as water. − In particular, systems involving water–hydrocarbon mixtures may exhibit phase behavior that is not fully captured by cubic EoS formulations.

Activity coefficient models such as UNIFAC are commonly used to represent liquid-phase nonidealities and phase separation behavior in systems involving water and hydrocarbons. These models are suitable for describing systems where strong immiscibility exists between phases and provide a practical representation of liquid-phase behavior. Moreover, being a predictive group-contribution method, UNIFAC does not require experimental binary interaction parameterswhich are unavailable for this systemmaking it well suited here. ,

Hybrid approaches such as the Predictive Soave–Redlich–Kwong (PSRK) model combine a cubic equation of state with a group-contribution activity-coefficient model and are typically employed for higher-pressure systems and where binary interaction parameters are unavailable. , Because the stripper operates near atmospheric pressure, where vapor-phase nonideality is negligible, the added complexity of a predictive equation of state is not warranted, and a direct activity-coefficient description of the liquid phase was preferred.

In this study, UNIFAC was used as the property method for the steam-stripping section because it captures the strong liquid-phase nonideality and the water–hydrocarbon phase separation that govern desorption, without requiring experimental binary parameters that are unavailable for this system. In the absence of measured multicomponent VLE data, the selected methods were assessed against physically expected behavior, the low solubility of permanent gases in a heavy hydrocarbon (Figure ) and the near-complete immiscibility of water and the hydrocarbon solvent, and the predicted stripping performance is benchmarked against the isoprene-recovery data of Feher et al.

Because the two sections present different phase behavior, section-specific property methods were used, PR for the absorption section and UNIFAC for the stripping section, as is standard in Aspen Plus.

3. Results and Discussion

3.1. Absorber

High efficiency in the separation of isoprene from the feed gas was demonstrated by the absorption column, with 97% of the isoprene being recovered into the solvent. Simultaneously, 99.8% of the air and 98% of the CO₂ were removed. To achieve this separation, a solvent-to-feed gas ratio of 1.2 was required. However, through effective recycling, net solvent loss was minimized to only 15 g kg^–1 isoprene, highlighting the sustainability of the process. A sensitivity analysis on solvent temperature, as presented in Figure , revealed a significant influence on the absorption efficiency of isoprene. It was found that the lowest energy requirement per kilogram of isoprene occurred at 2 °C. At lower temperatures, isoprene absorption efficiency was markedly increased. This enhancement was attributed to the high volatility of isoprene; reduced molecular energy at lower temperatures decreased isoprene’s tendency to remain in the gas phase, thereby promoting its dissolution into the liquid solvent phase. Conversely, a sharp decline in absorber efficiency was observed with increasing solvent temperature, reaching approximately 48% at room temperature. Similarly, lower solvent temperatures were associated with reduced energy requirements per kilogram of isoprene due to the higher yield achieved. Through multiple solvent recycling before stripping, we figured out at this temperature the maximum capacity of isoprene absorption is 9 g kg^–1 solvent.

4.

Effect of solvent temperature on absorber efficiency in isoprene separation (left axis) and energy demand per kg of isoprene produced in the final product (right axis). Simulation conditions are provided in Table .

The concentration of isoprene in the fermenter off-gas was determined by bacterial productivity and the airflow rate. A more concentrated isoprene stream was achieved under conditions of lower aeration combined with a higher bacterial metabolic rate. In this study, the effect of varying isoprene concentrations on solvent consumption, steam usage, and net energy demand was examined. The isoprene concentration was varied from 1 wt % to 4.3 wt %, while the lower explosive limit (LEL) of isoprene in air is 3.4 wt %. As illustrated in Figure , energy demand and utility consumption decrease sharply as the isoprene concentration in the feed gas increases. At 1 wt % isoprene, the process required about 26 MJ kg^–1 of product. A modest increase to 1.5 wt % reduced energy demand by 33% and lowered both steam and solvent usage by 34% (Figure ). At higher concentrations near the lower explosive limit (3.4 wt %), the energy requirement dropped further to approximately 10 MJ kg^–1, while steam usage fell to 0.5 kg per kilogram of isoprene. It would be possible to operate at concentrations above LEL, but the value was used for reference.

5.

Effect of feed gas isoprene concentration on energy demand per kg of isoprene produced (simulation conditions in Table ).

A comparison with the isoprene purification method reported by Feher et al. indicated that their system required approximately 1 kg of steam per kilogram of isoprene at an isoprene concentration of 1.8% v/v (∼4.3 wt %). Under similar conditions, improved efficiency was demonstrated by the present process, which required only 0.4 kg of steam per kilogram of isoprene. These findings highlight the potential for significant energy and resource savings through optimization of isoprene concentration in the fermenter off-gas.

3.2. Steam Stripping

In the steam stripping process, isoprene is extracted from a solvent. Efficiency of the steam stripping process is measured by the steam needed to strip 1 kg of isoprene. Figure shows how the solvent-to-feed gas ratio affects steam use for three feed gas flow rates: 800 kg h^–1, 1500 kg h^–1, and 3000 kg h^–1, at 1 wt % isoprene concentration. Steam use is lowest near a solvent-to-feed ratio of 1.2, requiring about 1.7 kg of steam per kg of isoprene. This reflects high energy efficiency. At lower ratios, too little solvent reduces isoprene capture and increases steam demand. On the contrary, excess solvent at higher ratios dilutes isoprene and raises steam use. A ratio of about 1.2 is the best condition for efficient separation.

Steam consumption in the stripping step is within a reasonable range for this process. Figure places the steam demand of the present process in the broader context of steam-stripping applications reported in the literature. Because these solutes differ markedly in polarity, volatility, and heteroatom content, Figure is intended as a qualitative overview of the typical range of steam requirements rather than a direct comparison. A controlled comparison is possible only for the same solute. In general, higher solute concentrations are expected to require less steam due to a greater driving force for mass transfer. However, the high volatility of isoprene facilitates its removal from the nonpolar solvent, even at relatively low concentrations. In the study by Feher et al., 2.8 kg of steam per kg of isoprene was required at a concentration of 0.76 wt %. In contrast, the current work demonstrates significantly improved efficiency, requiring only 1.7 kg of steam per kg of isoprene at 0.8 wt %. At a higher concentration of 2.4 wt %, steam consumption further decreased to 0.48 kg kg^–1 of isoprene. At this concentration, the proposed system achieved a 40% reduction in steam usage compared to the process reported by Feher et al.

8.

Steam requirement for steam stripping of various solutes reported in the literature. ,− The solute concentration refers to its weight percent in the solvent. The solutes differ widely in polarity, volatility, and heteroatom content; the figure is intended to illustrate the general range of steam demand across stripping applications rather than to provide a direct comparison.

3.3. Vacuum Degasser

The vacuum degasser serves to further remove air and CO₂ from the isoprene mixture. The effect of this unit on isoprene yield is illustrated in Figure . At lower pressures (i.e., under higher vacuum), more air and CO₂ are effectively removed, enabling approximately 99% of the isoprene entering the condenser to be recovered in liquid form. In contrast, at 1.5 atmthe pressure at which the stream exits the absorberapproximately 20% of the isoprene is lost in the vapor phase. This loss is attributed to the significantly greater retention of condensable gases at higher pressures, which can be up to eight times greater than that at lower pressures. The reduction in isoprene yield occurs because higher amounts of air and CO₂ dilute its partial pressure in the gas phase, lowering the driving force for condensation and making it less likely to reach the saturation pressure required for phase transition. Consequently, overall isoprene yield is diminished. Figure further demonstrates that, although operating under higher vacuum is inherently energy-intensive, it ultimately reduces the energy requirement per kilogram of isoprene produced.

9.

Effect of vacuum degasser pressure on mass flow rates of isoprene and noncondensable gases. “Isoprene produced” refers to the final liquid isoprene collected after condensation. “Isoprene (condenser)” denotes the isoprene vapor entering the condenser for phase change. “Air & CO₂ (condenser)” represents the combined flow of noncondensable gases entering the condenser (simulation conditions in Table ).

10.

Effect of vacuum pressure on net energy required per kilogram of isoprene produced (simulation conditions in Table ).

3.4. Energy Analysis

At 1% isoprene concentration in feed gas, 26 MJ of energy is needed to produce 1 kg of isoprene. Figure presents the breakdown of energy consumption across the major units within the isoprene recovery system. The solvent heater represents the largest energy consumer, accounting for 24.3% of the total energy input. The distribution indicates that thermal processesnamely heating, cooling, and steam generationdominate the system’s energy profile, collectively comprising over 60% of total consumption. This suggests that heat integration and thermal efficiency improvements represent the most promising areas for energy savings. Optimization strategies such as waste heat recovery, process intensification, and enhanced heat exchanger design may contribute significantly to reducing the overall energy footprint of the system while maintaining product yield and separation performance.

11.

Energy usage of different components for processing feed gas with 1 wt % isoprene (simulation conditions in Table ).

4. Conclusion

A comprehensive process simulation study was conducted to assess the technique and optimize energy consumption for purifying isoprene from a dilute air stream representative of microbial fermentation off-gas. The objective was to recover high-purity liquid isoprene while minimizing energy consumption and material losses. The separation scheme employed an integrated design comprising absorption, vacuum degassing, steam stripping, and condensation, with rigorous thermodynamic modeling guiding process decisions. The proposed system achieved 95% isoprene recovery with 98% product purity. Recycle loops for solvent, water, and residual isoprene were effectively incorporated to reduce material loss. The process demonstrated strong scalability and compatibility with biobased chemical production platforms, offering a practical solution for the efficient recovery of dilute volatile hydrocarbons.

Absorption was shown to be an effective method for isoprene separation from fermenter off-gas, particularly when integrated with heat recovery and recycling strategies. A key finding was the sensitivity of energy consumption to the isoprene concentration in the feed gas; enhanced aerobic fermentation processes that yield higher isoprene concentrations could significantly reduce energy requirements per kilogram of product.

It should be noted that the simulations relied on simplified thermodynamic models and steady-state operation, assuming ideal column performance and compensated solvent/water recycle loops. Dynamic effects, solvent degradation, and potential impurity accumulation were not considered. Furthermore, the purges were implemented as idealized component separations. The water purge is supported by water–hydrocarbon immiscibility (decantation); complete CO₂ removal is an approximation whose effect is negligible given the very small purged quantity. These necessary assumptions enabled tractable process design but highlight the need for future pilot-scale validation. Future work should focus on the development and operation of a pilot-scale system to validate the simulation results. Additionally, the design and selection of improved solvents may further lower energy demand and overall process costs as the major energy demand comes from the solvent heating. Overall, this study provides a clearer understanding of how feed concentration and process integration influence the efficiency of isoprene recovery, offering design principles that were not previously established.

The authors declare no competing financial interest.